mirror of
https://github.com/garrytan/gbrain.git
synced 2026-08-14 00:48:18 +00:00
Garry Tan
75fae742d5
Public, reproducible, cross-harness benchmark for the four ways agent memory fails (know-to-ask, push precision/recall, write-back fidelity, cross-session continuity). Hermetic PGLite, sealed gold, deterministic corpus, CI gate vs master's committed baseline, published JSON-schema foreign-runner contract. See docs/eval/BRAINBENCH.md.
22 lines
552 B
JSON
22 lines
552 B
JSON
{
|
|
"schema_version": 1,
|
|
"fixture_id": "gen-adv-012",
|
|
"suites": [
|
|
"know-to-ask"
|
|
],
|
|
"category": "adversarial",
|
|
"seed_pages": [
|
|
{
|
|
"slug": "people/petra-moonstead",
|
|
"content": "---\ntitle: Petra Moonstead\ntype: person\naliases: [petra]\nsummary: Founder of Bramblevolt Co.\n---\n\nPetra Moonstead is the founder of Bramblevolt Co. Petra shipped the v2 launch two weeks ahead of plan.\n"
|
|
}
|
|
],
|
|
"turns": [
|
|
{
|
|
"turn_id": 1,
|
|
"role": "user",
|
|
"text": "Did Petra Moonsteader reach out yet?"
|
|
}
|
|
]
|
|
}
|